1 /* $NetBSD: tcp_usrreq.c,v 1.142 2008/04/12 05:58:22 thorpej Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1997, 1998, 2005, 2006 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 38 * Facility, NASA Ames Research Center. 39 * This code is derived from software contributed to The NetBSD Foundation 40 * by Charles M. Hannum. 41 * This code is derived from software contributed to The NetBSD Foundation 42 * by Rui Paulo. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 3. All advertising materials mentioning features or use of this software 53 * must display the following acknowledgement: 54 * This product includes software developed by the NetBSD 55 * Foundation, Inc. and its contributors. 56 * 4. Neither the name of The NetBSD Foundation nor the names of its 57 * contributors may be used to endorse or promote products derived 58 * from this software without specific prior written permission. 59 * 60 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 61 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 62 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 63 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 64 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 65 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 66 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 67 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 68 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 69 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 70 * POSSIBILITY OF SUCH DAMAGE. 71 */ 72 73 /* 74 * Copyright (c) 1982, 1986, 1988, 1993, 1995 75 * The Regents of the University of California. All rights reserved. 76 * 77 * Redistribution and use in source and binary forms, with or without 78 * modification, are permitted provided that the following conditions 79 * are met: 80 * 1. Redistributions of source code must retain the above copyright 81 * notice, this list of conditions and the following disclaimer. 82 * 2. Redistributions in binary form must reproduce the above copyright 83 * notice, this list of conditions and the following disclaimer in the 84 * documentation and/or other materials provided with the distribution. 85 * 3. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)tcp_usrreq.c 8.5 (Berkeley) 6/21/95 102 */ 103 104 #include <sys/cdefs.h> 105 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.142 2008/04/12 05:58:22 thorpej Exp $"); 106 107 #include "opt_inet.h" 108 #include "opt_ipsec.h" 109 #include "opt_tcp_debug.h" 110 #include "opt_mbuftrace.h" 111 #include "rnd.h" 112 113 #include <sys/param.h> 114 #include <sys/systm.h> 115 #include <sys/kernel.h> 116 #include <sys/malloc.h> 117 #include <sys/mbuf.h> 118 #include <sys/socket.h> 119 #include <sys/socketvar.h> 120 #include <sys/protosw.h> 121 #include <sys/errno.h> 122 #include <sys/stat.h> 123 #include <sys/proc.h> 124 #include <sys/domain.h> 125 #include <sys/sysctl.h> 126 #include <sys/kauth.h> 127 128 #include <net/if.h> 129 #include <net/route.h> 130 131 #include <netinet/in.h> 132 #include <netinet/in_systm.h> 133 #include <netinet/in_var.h> 134 #include <netinet/ip.h> 135 #include <netinet/in_pcb.h> 136 #include <netinet/ip_var.h> 137 #include <netinet/in_offload.h> 138 139 #ifdef INET6 140 #ifndef INET 141 #include <netinet/in.h> 142 #endif 143 #include <netinet/ip6.h> 144 #include <netinet6/in6_pcb.h> 145 #include <netinet6/ip6_var.h> 146 #include <netinet6/scope6_var.h> 147 #endif 148 149 #include <netinet/tcp.h> 150 #include <netinet/tcp_fsm.h> 151 #include <netinet/tcp_seq.h> 152 #include <netinet/tcp_timer.h> 153 #include <netinet/tcp_var.h> 154 #include <netinet/tcp_private.h> 155 #include <netinet/tcp_congctl.h> 156 #include <netinet/tcpip.h> 157 #include <netinet/tcp_debug.h> 158 159 #include "opt_tcp_space.h" 160 161 #ifdef IPSEC 162 #include <netinet6/ipsec.h> 163 #endif /*IPSEC*/ 164 165 /* 166 * TCP protocol interface to socket abstraction. 167 */ 168 169 /* 170 * Process a TCP user request for TCP tb. If this is a send request 171 * then m is the mbuf chain of send data. If this is a timer expiration 172 * (called from the software clock routine), then timertype tells which timer. 173 */ 174 /*ARGSUSED*/ 175 int 176 tcp_usrreq(struct socket *so, int req, 177 struct mbuf *m, struct mbuf *nam, struct mbuf *control, struct lwp *l) 178 { 179 struct inpcb *inp; 180 #ifdef INET6 181 struct in6pcb *in6p; 182 #endif 183 struct tcpcb *tp = NULL; 184 int s; 185 int error = 0; 186 #ifdef TCP_DEBUG 187 int ostate = 0; 188 #endif 189 int family; /* family of the socket */ 190 191 family = so->so_proto->pr_domain->dom_family; 192 193 if (req == PRU_CONTROL) { 194 switch (family) { 195 #ifdef INET 196 case PF_INET: 197 return (in_control(so, (long)m, (void *)nam, 198 (struct ifnet *)control, l)); 199 #endif 200 #ifdef INET6 201 case PF_INET6: 202 return (in6_control(so, (long)m, (void *)nam, 203 (struct ifnet *)control, l)); 204 #endif 205 default: 206 return EAFNOSUPPORT; 207 } 208 } 209 210 s = splsoftnet(); 211 212 if (req == PRU_PURGEIF) { 213 switch (family) { 214 #ifdef INET 215 case PF_INET: 216 in_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 217 in_purgeif((struct ifnet *)control); 218 in_pcbpurgeif(&tcbtable, (struct ifnet *)control); 219 break; 220 #endif 221 #ifdef INET6 222 case PF_INET6: 223 in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 224 in6_purgeif((struct ifnet *)control); 225 in6_pcbpurgeif(&tcbtable, (struct ifnet *)control); 226 break; 227 #endif 228 default: 229 splx(s); 230 return (EAFNOSUPPORT); 231 } 232 splx(s); 233 return (0); 234 } 235 236 switch (family) { 237 #ifdef INET 238 case PF_INET: 239 inp = sotoinpcb(so); 240 #ifdef INET6 241 in6p = NULL; 242 #endif 243 break; 244 #endif 245 #ifdef INET6 246 case PF_INET6: 247 inp = NULL; 248 in6p = sotoin6pcb(so); 249 break; 250 #endif 251 default: 252 splx(s); 253 return EAFNOSUPPORT; 254 } 255 256 #ifdef DIAGNOSTIC 257 #ifdef INET6 258 if (inp && in6p) 259 panic("tcp_usrreq: both inp and in6p set to non-NULL"); 260 #endif 261 if (req != PRU_SEND && req != PRU_SENDOOB && control) 262 panic("tcp_usrreq: unexpected control mbuf"); 263 #endif 264 /* 265 * When a TCP is attached to a socket, then there will be 266 * a (struct inpcb) pointed at by the socket, and this 267 * structure will point at a subsidary (struct tcpcb). 268 */ 269 #ifndef INET6 270 if (inp == 0 && req != PRU_ATTACH) 271 #else 272 if ((inp == 0 && in6p == 0) && req != PRU_ATTACH) 273 #endif 274 { 275 error = EINVAL; 276 goto release; 277 } 278 #ifdef INET 279 if (inp) { 280 tp = intotcpcb(inp); 281 /* WHAT IF TP IS 0? */ 282 #ifdef KPROF 283 tcp_acounts[tp->t_state][req]++; 284 #endif 285 #ifdef TCP_DEBUG 286 ostate = tp->t_state; 287 #endif 288 } 289 #endif 290 #ifdef INET6 291 if (in6p) { 292 tp = in6totcpcb(in6p); 293 /* WHAT IF TP IS 0? */ 294 #ifdef KPROF 295 tcp_acounts[tp->t_state][req]++; 296 #endif 297 #ifdef TCP_DEBUG 298 ostate = tp->t_state; 299 #endif 300 } 301 #endif 302 303 switch (req) { 304 305 /* 306 * TCP attaches to socket via PRU_ATTACH, reserving space, 307 * and an internet control block. 308 */ 309 case PRU_ATTACH: 310 #ifndef INET6 311 if (inp != 0) 312 #else 313 if (inp != 0 || in6p != 0) 314 #endif 315 { 316 error = EISCONN; 317 break; 318 } 319 error = tcp_attach(so); 320 if (error) 321 break; 322 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 323 so->so_linger = TCP_LINGERTIME; 324 tp = sototcpcb(so); 325 break; 326 327 /* 328 * PRU_DETACH detaches the TCP protocol from the socket. 329 */ 330 case PRU_DETACH: 331 tp = tcp_disconnect(tp); 332 break; 333 334 /* 335 * Give the socket an address. 336 */ 337 case PRU_BIND: 338 switch (family) { 339 #ifdef INET 340 case PF_INET: 341 error = in_pcbbind(inp, nam, l); 342 break; 343 #endif 344 #ifdef INET6 345 case PF_INET6: 346 error = in6_pcbbind(in6p, nam, l); 347 if (!error) { 348 /* mapped addr case */ 349 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) 350 tp->t_family = AF_INET; 351 else 352 tp->t_family = AF_INET6; 353 } 354 break; 355 #endif 356 } 357 break; 358 359 /* 360 * Prepare to accept connections. 361 */ 362 case PRU_LISTEN: 363 #ifdef INET 364 if (inp && inp->inp_lport == 0) { 365 error = in_pcbbind(inp, (struct mbuf *)0, l); 366 if (error) 367 break; 368 } 369 #endif 370 #ifdef INET6 371 if (in6p && in6p->in6p_lport == 0) { 372 error = in6_pcbbind(in6p, (struct mbuf *)0, 373 (struct lwp *)0); 374 if (error) 375 break; 376 } 377 #endif 378 tp->t_state = TCPS_LISTEN; 379 break; 380 381 /* 382 * Initiate connection to peer. 383 * Create a template for use in transmissions on this connection. 384 * Enter SYN_SENT state, and mark socket as connecting. 385 * Start keep-alive timer, and seed output sequence space. 386 * Send initial segment on connection. 387 */ 388 case PRU_CONNECT: 389 #ifdef INET 390 if (inp) { 391 if (inp->inp_lport == 0) { 392 error = in_pcbbind(inp, (struct mbuf *)0, l); 393 if (error) 394 break; 395 } 396 error = in_pcbconnect(inp, nam, l); 397 } 398 #endif 399 #ifdef INET6 400 if (in6p) { 401 if (in6p->in6p_lport == 0) { 402 error = in6_pcbbind(in6p, (struct mbuf *)0, 403 (struct lwp *)0); 404 if (error) 405 break; 406 } 407 error = in6_pcbconnect(in6p, nam, l); 408 if (!error) { 409 /* mapped addr case */ 410 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 411 tp->t_family = AF_INET; 412 else 413 tp->t_family = AF_INET6; 414 } 415 } 416 #endif 417 if (error) 418 break; 419 tp->t_template = tcp_template(tp); 420 if (tp->t_template == 0) { 421 #ifdef INET 422 if (inp) 423 in_pcbdisconnect(inp); 424 #endif 425 #ifdef INET6 426 if (in6p) 427 in6_pcbdisconnect(in6p); 428 #endif 429 error = ENOBUFS; 430 break; 431 } 432 /* 433 * Compute window scaling to request. 434 * XXX: This should be moved to tcp_output(). 435 */ 436 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 437 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 438 tp->request_r_scale++; 439 soisconnecting(so); 440 TCP_STATINC(TCP_STAT_CONNATTEMPT); 441 tp->t_state = TCPS_SYN_SENT; 442 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 443 tp->iss = tcp_new_iss(tp, 0); 444 tcp_sendseqinit(tp); 445 error = tcp_output(tp); 446 break; 447 448 /* 449 * Create a TCP connection between two sockets. 450 */ 451 case PRU_CONNECT2: 452 error = EOPNOTSUPP; 453 break; 454 455 /* 456 * Initiate disconnect from peer. 457 * If connection never passed embryonic stage, just drop; 458 * else if don't need to let data drain, then can just drop anyways, 459 * else have to begin TCP shutdown process: mark socket disconnecting, 460 * drain unread data, state switch to reflect user close, and 461 * send segment (e.g. FIN) to peer. Socket will be really disconnected 462 * when peer sends FIN and acks ours. 463 * 464 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 465 */ 466 case PRU_DISCONNECT: 467 tp = tcp_disconnect(tp); 468 break; 469 470 /* 471 * Accept a connection. Essentially all the work is 472 * done at higher levels; just return the address 473 * of the peer, storing through addr. 474 */ 475 case PRU_ACCEPT: 476 #ifdef INET 477 if (inp) 478 in_setpeeraddr(inp, nam); 479 #endif 480 #ifdef INET6 481 if (in6p) 482 in6_setpeeraddr(in6p, nam); 483 #endif 484 break; 485 486 /* 487 * Mark the connection as being incapable of further output. 488 */ 489 case PRU_SHUTDOWN: 490 socantsendmore(so); 491 tp = tcp_usrclosed(tp); 492 if (tp) 493 error = tcp_output(tp); 494 break; 495 496 /* 497 * After a receive, possibly send window update to peer. 498 */ 499 case PRU_RCVD: 500 /* 501 * soreceive() calls this function when a user receives 502 * ancillary data on a listening socket. We don't call 503 * tcp_output in such a case, since there is no header 504 * template for a listening socket and hence the kernel 505 * will panic. 506 */ 507 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0) 508 (void) tcp_output(tp); 509 break; 510 511 /* 512 * Do a send by putting data in output queue and updating urgent 513 * marker if URG set. Possibly send more data. 514 */ 515 case PRU_SEND: 516 if (control && control->m_len) { 517 m_freem(control); 518 m_freem(m); 519 error = EINVAL; 520 break; 521 } 522 sbappendstream(&so->so_snd, m); 523 error = tcp_output(tp); 524 break; 525 526 /* 527 * Abort the TCP. 528 */ 529 case PRU_ABORT: 530 tp = tcp_drop(tp, ECONNABORTED); 531 break; 532 533 case PRU_SENSE: 534 /* 535 * stat: don't bother with a blocksize. 536 */ 537 splx(s); 538 return (0); 539 540 case PRU_RCVOOB: 541 if (control && control->m_len) { 542 m_freem(control); 543 m_freem(m); 544 error = EINVAL; 545 break; 546 } 547 if ((so->so_oobmark == 0 && 548 (so->so_state & SS_RCVATMARK) == 0) || 549 so->so_options & SO_OOBINLINE || 550 tp->t_oobflags & TCPOOB_HADDATA) { 551 error = EINVAL; 552 break; 553 } 554 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 555 error = EWOULDBLOCK; 556 break; 557 } 558 m->m_len = 1; 559 *mtod(m, char *) = tp->t_iobc; 560 if (((long)nam & MSG_PEEK) == 0) 561 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 562 break; 563 564 case PRU_SENDOOB: 565 if (sbspace(&so->so_snd) < -512) { 566 m_freem(m); 567 error = ENOBUFS; 568 break; 569 } 570 /* 571 * According to RFC961 (Assigned Protocols), 572 * the urgent pointer points to the last octet 573 * of urgent data. We continue, however, 574 * to consider it to indicate the first octet 575 * of data past the urgent section. 576 * Otherwise, snd_up should be one lower. 577 */ 578 sbappendstream(&so->so_snd, m); 579 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 580 tp->t_force = 1; 581 error = tcp_output(tp); 582 tp->t_force = 0; 583 break; 584 585 case PRU_SOCKADDR: 586 #ifdef INET 587 if (inp) 588 in_setsockaddr(inp, nam); 589 #endif 590 #ifdef INET6 591 if (in6p) 592 in6_setsockaddr(in6p, nam); 593 #endif 594 break; 595 596 case PRU_PEERADDR: 597 #ifdef INET 598 if (inp) 599 in_setpeeraddr(inp, nam); 600 #endif 601 #ifdef INET6 602 if (in6p) 603 in6_setpeeraddr(in6p, nam); 604 #endif 605 break; 606 607 default: 608 panic("tcp_usrreq"); 609 } 610 #ifdef TCP_DEBUG 611 if (tp && (so->so_options & SO_DEBUG)) 612 tcp_trace(TA_USER, ostate, tp, NULL, req); 613 #endif 614 615 release: 616 splx(s); 617 return (error); 618 } 619 620 static void 621 change_keepalive(struct socket *so, struct tcpcb *tp) 622 { 623 tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl; 624 TCP_TIMER_DISARM(tp, TCPT_KEEP); 625 TCP_TIMER_DISARM(tp, TCPT_2MSL); 626 627 if (tp->t_state == TCPS_SYN_RECEIVED || 628 tp->t_state == TCPS_SYN_SENT) { 629 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 630 } else if (so->so_options & SO_KEEPALIVE && 631 tp->t_state <= TCPS_CLOSE_WAIT) { 632 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl); 633 } else { 634 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 635 } 636 637 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 638 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 639 } 640 641 642 int 643 tcp_ctloutput(int op, struct socket *so, int level, int optname, 644 struct mbuf **mp) 645 { 646 int error = 0, s; 647 struct inpcb *inp; 648 #ifdef INET6 649 struct in6pcb *in6p; 650 #endif 651 struct tcpcb *tp; 652 struct mbuf *m; 653 int i; 654 u_int ui; 655 int family; /* family of the socket */ 656 657 family = so->so_proto->pr_domain->dom_family; 658 659 s = splsoftnet(); 660 switch (family) { 661 #ifdef INET 662 case PF_INET: 663 inp = sotoinpcb(so); 664 #ifdef INET6 665 in6p = NULL; 666 #endif 667 break; 668 #endif 669 #ifdef INET6 670 case PF_INET6: 671 inp = NULL; 672 in6p = sotoin6pcb(so); 673 break; 674 #endif 675 default: 676 splx(s); 677 panic("%s: af %d", __func__, family); 678 } 679 #ifndef INET6 680 if (inp == NULL) 681 #else 682 if (inp == NULL && in6p == NULL) 683 #endif 684 { 685 splx(s); 686 if (op == PRCO_SETOPT && *mp) 687 (void) m_free(*mp); 688 return (ECONNRESET); 689 } 690 if (level != IPPROTO_TCP) { 691 switch (family) { 692 #ifdef INET 693 case PF_INET: 694 error = ip_ctloutput(op, so, level, optname, mp); 695 break; 696 #endif 697 #ifdef INET6 698 case PF_INET6: 699 error = ip6_ctloutput(op, so, level, optname, mp); 700 break; 701 #endif 702 } 703 splx(s); 704 return (error); 705 } 706 if (inp) 707 tp = intotcpcb(inp); 708 #ifdef INET6 709 else if (in6p) 710 tp = in6totcpcb(in6p); 711 #endif 712 else 713 tp = NULL; 714 715 switch (op) { 716 717 case PRCO_SETOPT: 718 m = *mp; 719 switch (optname) { 720 721 #ifdef TCP_SIGNATURE 722 case TCP_MD5SIG: 723 if (m == NULL || m->m_len != sizeof(int)) 724 error = EINVAL; 725 if (error) 726 break; 727 if (*mtod(m, int *) > 0) 728 tp->t_flags |= TF_SIGNATURE; 729 else 730 tp->t_flags &= ~TF_SIGNATURE; 731 break; 732 #endif /* TCP_SIGNATURE */ 733 734 case TCP_NODELAY: 735 if (m == NULL || m->m_len != sizeof(int)) 736 error = EINVAL; 737 else if (*mtod(m, int *)) 738 tp->t_flags |= TF_NODELAY; 739 else 740 tp->t_flags &= ~TF_NODELAY; 741 break; 742 743 case TCP_MAXSEG: 744 if (m && m->m_len == sizeof(int) && 745 (i = *mtod(m, int *)) > 0 && 746 i <= tp->t_peermss) 747 tp->t_peermss = i; /* limit on send size */ 748 else 749 error = EINVAL; 750 break; 751 #ifdef notyet 752 case TCP_CONGCTL: 753 if (m == NULL) 754 error = EINVAL; 755 error = tcp_congctl_select(tp, mtod(m, char *)); 756 #endif 757 break; 758 759 case TCP_KEEPIDLE: 760 if (m && m->m_len == sizeof(u_int) && 761 (ui = *mtod(m, u_int *)) > 0) { 762 tp->t_keepidle = ui; 763 change_keepalive(so, tp); 764 } else 765 error = EINVAL; 766 break; 767 768 case TCP_KEEPINTVL: 769 if (m && m->m_len == sizeof(u_int) && 770 (ui = *mtod(m, u_int *)) > 0) { 771 tp->t_keepintvl = ui; 772 change_keepalive(so, tp); 773 } else 774 error = EINVAL; 775 break; 776 777 case TCP_KEEPCNT: 778 if (m && m->m_len == sizeof(u_int) && 779 (ui = *mtod(m, u_int *)) > 0) { 780 tp->t_keepcnt = ui; 781 change_keepalive(so, tp); 782 } else 783 error = EINVAL; 784 break; 785 786 case TCP_KEEPINIT: 787 if (m && m->m_len == sizeof(u_int) && 788 (ui = *mtod(m, u_int *)) > 0) { 789 tp->t_keepinit = ui; 790 change_keepalive(so, tp); 791 } else 792 error = EINVAL; 793 break; 794 795 default: 796 error = ENOPROTOOPT; 797 break; 798 } 799 if (m) 800 (void) m_free(m); 801 break; 802 803 case PRCO_GETOPT: 804 *mp = m = m_intopt(so, 0); 805 806 switch (optname) { 807 #ifdef TCP_SIGNATURE 808 case TCP_MD5SIG: 809 *mtod(m, int *) = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 810 break; 811 #endif 812 case TCP_NODELAY: 813 *mtod(m, int *) = tp->t_flags & TF_NODELAY; 814 break; 815 case TCP_MAXSEG: 816 *mtod(m, int *) = tp->t_peermss; 817 break; 818 #ifdef notyet 819 case TCP_CONGCTL: 820 break; 821 #endif 822 default: 823 error = ENOPROTOOPT; 824 break; 825 } 826 break; 827 } 828 splx(s); 829 return (error); 830 } 831 832 #ifndef TCP_SENDSPACE 833 #define TCP_SENDSPACE 1024*32 834 #endif 835 int tcp_sendspace = TCP_SENDSPACE; 836 #ifndef TCP_RECVSPACE 837 #define TCP_RECVSPACE 1024*32 838 #endif 839 int tcp_recvspace = TCP_RECVSPACE; 840 841 /* 842 * Attach TCP protocol to socket, allocating 843 * internet protocol control block, tcp control block, 844 * bufer space, and entering LISTEN state if to accept connections. 845 */ 846 int 847 tcp_attach(struct socket *so) 848 { 849 struct tcpcb *tp; 850 struct inpcb *inp; 851 #ifdef INET6 852 struct in6pcb *in6p; 853 #endif 854 int error; 855 int family; /* family of the socket */ 856 857 family = so->so_proto->pr_domain->dom_family; 858 859 #ifdef MBUFTRACE 860 so->so_mowner = &tcp_sock_mowner; 861 so->so_rcv.sb_mowner = &tcp_sock_rx_mowner; 862 so->so_snd.sb_mowner = &tcp_sock_tx_mowner; 863 #endif 864 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 865 error = soreserve(so, tcp_sendspace, tcp_recvspace); 866 if (error) 867 return (error); 868 } 869 870 so->so_rcv.sb_flags |= SB_AUTOSIZE; 871 so->so_snd.sb_flags |= SB_AUTOSIZE; 872 873 switch (family) { 874 #ifdef INET 875 case PF_INET: 876 error = in_pcballoc(so, &tcbtable); 877 if (error) 878 return (error); 879 inp = sotoinpcb(so); 880 #ifdef INET6 881 in6p = NULL; 882 #endif 883 break; 884 #endif 885 #ifdef INET6 886 case PF_INET6: 887 error = in6_pcballoc(so, &tcbtable); 888 if (error) 889 return (error); 890 inp = NULL; 891 in6p = sotoin6pcb(so); 892 break; 893 #endif 894 default: 895 return EAFNOSUPPORT; 896 } 897 if (inp) 898 tp = tcp_newtcpcb(family, (void *)inp); 899 #ifdef INET6 900 else if (in6p) 901 tp = tcp_newtcpcb(family, (void *)in6p); 902 #endif 903 else 904 tp = NULL; 905 906 if (tp == 0) { 907 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 908 909 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 910 #ifdef INET 911 if (inp) 912 in_pcbdetach(inp); 913 #endif 914 #ifdef INET6 915 if (in6p) 916 in6_pcbdetach(in6p); 917 #endif 918 so->so_state |= nofd; 919 return (ENOBUFS); 920 } 921 tp->t_state = TCPS_CLOSED; 922 return (0); 923 } 924 925 /* 926 * Initiate (or continue) disconnect. 927 * If embryonic state, just send reset (once). 928 * If in ``let data drain'' option and linger null, just drop. 929 * Otherwise (hard), mark socket disconnecting and drop 930 * current input data; switch states based on user close, and 931 * send segment to peer (with FIN). 932 */ 933 struct tcpcb * 934 tcp_disconnect(struct tcpcb *tp) 935 { 936 struct socket *so; 937 938 if (tp->t_inpcb) 939 so = tp->t_inpcb->inp_socket; 940 #ifdef INET6 941 else if (tp->t_in6pcb) 942 so = tp->t_in6pcb->in6p_socket; 943 #endif 944 else 945 so = NULL; 946 947 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 948 tp = tcp_close(tp); 949 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 950 tp = tcp_drop(tp, 0); 951 else { 952 soisdisconnecting(so); 953 sbflush(&so->so_rcv); 954 tp = tcp_usrclosed(tp); 955 if (tp) 956 (void) tcp_output(tp); 957 } 958 return (tp); 959 } 960 961 /* 962 * User issued close, and wish to trail through shutdown states: 963 * if never received SYN, just forget it. If got a SYN from peer, 964 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 965 * If already got a FIN from peer, then almost done; go to LAST_ACK 966 * state. In all other cases, have already sent FIN to peer (e.g. 967 * after PRU_SHUTDOWN), and just have to play tedious game waiting 968 * for peer to send FIN or not respond to keep-alives, etc. 969 * We can let the user exit from the close as soon as the FIN is acked. 970 */ 971 struct tcpcb * 972 tcp_usrclosed(struct tcpcb *tp) 973 { 974 975 switch (tp->t_state) { 976 977 case TCPS_CLOSED: 978 case TCPS_LISTEN: 979 case TCPS_SYN_SENT: 980 tp->t_state = TCPS_CLOSED; 981 tp = tcp_close(tp); 982 break; 983 984 case TCPS_SYN_RECEIVED: 985 case TCPS_ESTABLISHED: 986 tp->t_state = TCPS_FIN_WAIT_1; 987 break; 988 989 case TCPS_CLOSE_WAIT: 990 tp->t_state = TCPS_LAST_ACK; 991 break; 992 } 993 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 994 struct socket *so; 995 if (tp->t_inpcb) 996 so = tp->t_inpcb->inp_socket; 997 #ifdef INET6 998 else if (tp->t_in6pcb) 999 so = tp->t_in6pcb->in6p_socket; 1000 #endif 1001 else 1002 so = NULL; 1003 if (so) 1004 soisdisconnected(so); 1005 /* 1006 * If we are in FIN_WAIT_2, we arrived here because the 1007 * application did a shutdown of the send side. Like the 1008 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after 1009 * a full close, we start a timer to make sure sockets are 1010 * not left in FIN_WAIT_2 forever. 1011 */ 1012 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 1013 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 1014 } 1015 return (tp); 1016 } 1017 1018 /* 1019 * sysctl helper routine for net.inet.ip.mssdflt. it can't be less 1020 * than 32. 1021 */ 1022 static int 1023 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS) 1024 { 1025 int error, mssdflt; 1026 struct sysctlnode node; 1027 1028 mssdflt = tcp_mssdflt; 1029 node = *rnode; 1030 node.sysctl_data = &mssdflt; 1031 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1032 if (error || newp == NULL) 1033 return (error); 1034 1035 if (mssdflt < 32) 1036 return (EINVAL); 1037 tcp_mssdflt = mssdflt; 1038 1039 return (0); 1040 } 1041 1042 /* 1043 * sysctl helper routine for setting port related values under 1044 * net.inet.ip and net.inet6.ip6. does basic range checking and does 1045 * additional checks for each type. this code has placed in 1046 * tcp_input.c since INET and INET6 both use the same tcp code. 1047 * 1048 * this helper is not static so that both inet and inet6 can use it. 1049 */ 1050 int 1051 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS) 1052 { 1053 int error, tmp; 1054 int apmin, apmax; 1055 #ifndef IPNOPRIVPORTS 1056 int lpmin, lpmax; 1057 #endif /* IPNOPRIVPORTS */ 1058 struct sysctlnode node; 1059 1060 if (namelen != 0) 1061 return (EINVAL); 1062 1063 switch (name[-3]) { 1064 #ifdef INET 1065 case PF_INET: 1066 apmin = anonportmin; 1067 apmax = anonportmax; 1068 #ifndef IPNOPRIVPORTS 1069 lpmin = lowportmin; 1070 lpmax = lowportmax; 1071 #endif /* IPNOPRIVPORTS */ 1072 break; 1073 #endif /* INET */ 1074 #ifdef INET6 1075 case PF_INET6: 1076 apmin = ip6_anonportmin; 1077 apmax = ip6_anonportmax; 1078 #ifndef IPNOPRIVPORTS 1079 lpmin = ip6_lowportmin; 1080 lpmax = ip6_lowportmax; 1081 #endif /* IPNOPRIVPORTS */ 1082 break; 1083 #endif /* INET6 */ 1084 default: 1085 return (EINVAL); 1086 } 1087 1088 /* 1089 * insert temporary copy into node, perform lookup on 1090 * temporary, then restore pointer 1091 */ 1092 node = *rnode; 1093 tmp = *(int*)rnode->sysctl_data; 1094 node.sysctl_data = &tmp; 1095 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1096 if (error || newp == NULL) 1097 return (error); 1098 1099 /* 1100 * simple port range check 1101 */ 1102 if (tmp < 0 || tmp > 65535) 1103 return (EINVAL); 1104 1105 /* 1106 * per-node range checks 1107 */ 1108 switch (rnode->sysctl_num) { 1109 case IPCTL_ANONPORTMIN: 1110 if (tmp >= apmax) 1111 return (EINVAL); 1112 #ifndef IPNOPRIVPORTS 1113 if (tmp < IPPORT_RESERVED) 1114 return (EINVAL); 1115 #endif /* IPNOPRIVPORTS */ 1116 break; 1117 1118 case IPCTL_ANONPORTMAX: 1119 if (apmin >= tmp) 1120 return (EINVAL); 1121 #ifndef IPNOPRIVPORTS 1122 if (tmp < IPPORT_RESERVED) 1123 return (EINVAL); 1124 #endif /* IPNOPRIVPORTS */ 1125 break; 1126 1127 #ifndef IPNOPRIVPORTS 1128 case IPCTL_LOWPORTMIN: 1129 if (tmp >= lpmax || 1130 tmp > IPPORT_RESERVEDMAX || 1131 tmp < IPPORT_RESERVEDMIN) 1132 return (EINVAL); 1133 break; 1134 1135 case IPCTL_LOWPORTMAX: 1136 if (lpmin >= tmp || 1137 tmp > IPPORT_RESERVEDMAX || 1138 tmp < IPPORT_RESERVEDMIN) 1139 return (EINVAL); 1140 break; 1141 #endif /* IPNOPRIVPORTS */ 1142 1143 default: 1144 return (EINVAL); 1145 } 1146 1147 *(int*)rnode->sysctl_data = tmp; 1148 1149 return (0); 1150 } 1151 1152 /* 1153 * The superuser can drop any connection. Normal users can only drop 1154 * their own connections. 1155 */ 1156 static inline int 1157 check_sockuid(struct socket *sockp, kauth_cred_t cred) 1158 { 1159 uid_t sockuid; 1160 1161 sockuid = sockp->so_uidinfo->ui_uid; 1162 if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) == 0 || 1163 sockuid == kauth_cred_getuid(cred) || 1164 sockuid == kauth_cred_geteuid(cred)) 1165 return 0; 1166 return EACCES; 1167 } 1168 1169 static inline int 1170 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp) 1171 { 1172 size_t sz; 1173 int error; 1174 uid_t uid; 1175 1176 uid = sockp->so_uidinfo->ui_uid; 1177 if (oldp) { 1178 sz = MIN(sizeof(uid), *oldlenp); 1179 error = copyout(&uid, oldp, sz); 1180 if (error) 1181 return error; 1182 } 1183 *oldlenp = sizeof(uid); 1184 return 0; 1185 } 1186 1187 static inline int 1188 inet4_ident_core(struct in_addr raddr, u_int rport, 1189 struct in_addr laddr, u_int lport, 1190 void *oldp, size_t *oldlenp, 1191 struct lwp *l, int dodrop) 1192 { 1193 struct inpcb *inp; 1194 struct socket *sockp; 1195 1196 inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport); 1197 1198 if (inp == NULL || (sockp = inp->inp_socket) == NULL) 1199 return ESRCH; 1200 1201 if (dodrop) { 1202 struct tcpcb *tp; 1203 1204 if (inp == NULL || (tp = intotcpcb(inp)) == NULL || 1205 (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0) 1206 return ESRCH; 1207 1208 if (check_sockuid(inp->inp_socket, l->l_cred) != 0) 1209 return EACCES; 1210 1211 (void)tcp_drop(tp, ECONNABORTED); 1212 return 0; 1213 } 1214 else 1215 return copyout_uid(sockp, oldp, oldlenp); 1216 } 1217 1218 #ifdef INET6 1219 static inline int 1220 inet6_ident_core(struct in6_addr *raddr, u_int rport, 1221 struct in6_addr *laddr, u_int lport, 1222 void *oldp, size_t *oldlenp, 1223 struct lwp *l, int dodrop) 1224 { 1225 struct in6pcb *in6p; 1226 struct socket *sockp; 1227 1228 in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0); 1229 1230 if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL) 1231 return ESRCH; 1232 1233 if (dodrop) { 1234 struct tcpcb *tp; 1235 1236 if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL || 1237 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0) 1238 return ESRCH; 1239 1240 if (check_sockuid(in6p->in6p_socket, l->l_cred) != 0) 1241 return EACCES; 1242 1243 (void)tcp_drop(tp, ECONNABORTED); 1244 return 0; 1245 } 1246 else 1247 return copyout_uid(sockp, oldp, oldlenp); 1248 } 1249 #endif 1250 1251 /* 1252 * sysctl helper routine for the net.inet.tcp.drop and 1253 * net.inet6.tcp6.drop nodes. 1254 */ 1255 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident 1256 1257 /* 1258 * sysctl helper routine for the net.inet.tcp.ident and 1259 * net.inet6.tcp6.ident nodes. contains backwards compat code for the 1260 * old way of looking up the ident information for ipv4 which involves 1261 * stuffing the port/addr pairs into the mib lookup. 1262 */ 1263 static int 1264 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS) 1265 { 1266 #ifdef INET 1267 struct sockaddr_in *si4[2]; 1268 #endif /* INET */ 1269 #ifdef INET6 1270 struct sockaddr_in6 *si6[2]; 1271 #endif /* INET6 */ 1272 struct sockaddr_storage sa[2]; 1273 int error, pf, dodrop, s; 1274 1275 dodrop = name[-1] == TCPCTL_DROP; 1276 if (dodrop) { 1277 if (oldp != NULL || *oldlenp != 0) 1278 return EINVAL; 1279 if (newp == NULL) 1280 return EPERM; 1281 if (newlen < sizeof(sa)) 1282 return ENOMEM; 1283 } 1284 if (namelen != 4 && namelen != 0) 1285 return EINVAL; 1286 if (name[-2] != IPPROTO_TCP) 1287 return EINVAL; 1288 pf = name[-3]; 1289 1290 /* old style lookup, ipv4 only */ 1291 if (namelen == 4) { 1292 #ifdef INET 1293 struct in_addr laddr, raddr; 1294 u_int lport, rport; 1295 1296 if (pf != PF_INET) 1297 return EPROTONOSUPPORT; 1298 raddr.s_addr = (uint32_t)name[0]; 1299 rport = (u_int)name[1]; 1300 laddr.s_addr = (uint32_t)name[2]; 1301 lport = (u_int)name[3]; 1302 1303 s = splsoftnet(); 1304 error = inet4_ident_core(raddr, rport, laddr, lport, 1305 oldp, oldlenp, l, dodrop); 1306 splx(s); 1307 return error; 1308 #else /* INET */ 1309 return EINVAL; 1310 #endif /* INET */ 1311 } 1312 1313 if (newp == NULL || newlen != sizeof(sa)) 1314 return EINVAL; 1315 error = copyin(newp, &sa, newlen); 1316 if (error) 1317 return error; 1318 1319 /* 1320 * requested families must match 1321 */ 1322 if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family) 1323 return EINVAL; 1324 1325 switch (pf) { 1326 #ifdef INET6 1327 case PF_INET6: 1328 si6[0] = (struct sockaddr_in6*)&sa[0]; 1329 si6[1] = (struct sockaddr_in6*)&sa[1]; 1330 if (si6[0]->sin6_len != sizeof(*si6[0]) || 1331 si6[1]->sin6_len != sizeof(*si6[1])) 1332 return EINVAL; 1333 1334 if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) && 1335 !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) { 1336 error = sa6_embedscope(si6[0], ip6_use_defzone); 1337 if (error) 1338 return error; 1339 error = sa6_embedscope(si6[1], ip6_use_defzone); 1340 if (error) 1341 return error; 1342 1343 s = splsoftnet(); 1344 error = inet6_ident_core(&si6[0]->sin6_addr, 1345 si6[0]->sin6_port, &si6[1]->sin6_addr, 1346 si6[1]->sin6_port, oldp, oldlenp, l, dodrop); 1347 splx(s); 1348 return error; 1349 } 1350 1351 if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) != 1352 IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) 1353 return EINVAL; 1354 1355 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]); 1356 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]); 1357 /*FALLTHROUGH*/ 1358 #endif /* INET6 */ 1359 #ifdef INET 1360 case PF_INET: 1361 si4[0] = (struct sockaddr_in*)&sa[0]; 1362 si4[1] = (struct sockaddr_in*)&sa[1]; 1363 if (si4[0]->sin_len != sizeof(*si4[0]) || 1364 si4[0]->sin_len != sizeof(*si4[1])) 1365 return EINVAL; 1366 1367 s = splsoftnet(); 1368 error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port, 1369 si4[1]->sin_addr, si4[1]->sin_port, 1370 oldp, oldlenp, l, dodrop); 1371 splx(s); 1372 return error; 1373 #endif /* INET */ 1374 default: 1375 return EPROTONOSUPPORT; 1376 } 1377 } 1378 1379 /* 1380 * sysctl helper for the inet and inet6 pcblists. handles tcp/udp and 1381 * inet/inet6, as well as raw pcbs for each. specifically not 1382 * declared static so that raw sockets and udp/udp6 can use it as 1383 * well. 1384 */ 1385 int 1386 sysctl_inpcblist(SYSCTLFN_ARGS) 1387 { 1388 #ifdef INET 1389 struct sockaddr_in *in; 1390 const struct inpcb *inp; 1391 #endif 1392 #ifdef INET6 1393 struct sockaddr_in6 *in6; 1394 const struct in6pcb *in6p; 1395 #endif 1396 /* 1397 * sysctl_data is const, but CIRCLEQ_FOREACH can't use a const 1398 * struct inpcbtable pointer, so we have to discard const. :-/ 1399 */ 1400 struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data); 1401 const struct inpcb_hdr *inph; 1402 struct tcpcb *tp; 1403 struct kinfo_pcb pcb; 1404 char *dp; 1405 u_int op, arg; 1406 size_t len, needed, elem_size, out_size; 1407 int error, elem_count, pf, proto, pf2; 1408 1409 if (namelen != 4) 1410 return (EINVAL); 1411 1412 if (oldp != NULL) { 1413 len = *oldlenp; 1414 elem_size = name[2]; 1415 elem_count = name[3]; 1416 if (elem_size != sizeof(pcb)) 1417 return EINVAL; 1418 } else { 1419 len = 0; 1420 elem_count = INT_MAX; 1421 elem_size = sizeof(pcb); 1422 } 1423 error = 0; 1424 dp = oldp; 1425 op = name[0]; 1426 arg = name[1]; 1427 out_size = elem_size; 1428 needed = 0; 1429 1430 if (namelen == 1 && name[0] == CTL_QUERY) 1431 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1432 1433 if (name - oname != 4) 1434 return (EINVAL); 1435 1436 pf = oname[1]; 1437 proto = oname[2]; 1438 pf2 = (oldp != NULL) ? pf : 0; 1439 1440 CIRCLEQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) { 1441 #ifdef INET 1442 inp = (const struct inpcb *)inph; 1443 #endif 1444 #ifdef INET6 1445 in6p = (const struct in6pcb *)inph; 1446 #endif 1447 1448 if (inph->inph_af != pf) 1449 continue; 1450 1451 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1452 KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL, 1453 NULL) != 0) 1454 continue; 1455 1456 memset(&pcb, 0, sizeof(pcb)); 1457 1458 pcb.ki_family = pf; 1459 pcb.ki_type = proto; 1460 1461 switch (pf2) { 1462 case 0: 1463 /* just probing for size */ 1464 break; 1465 #ifdef INET 1466 case PF_INET: 1467 pcb.ki_family = inp->inp_socket->so_proto-> 1468 pr_domain->dom_family; 1469 pcb.ki_type = inp->inp_socket->so_proto-> 1470 pr_type; 1471 pcb.ki_protocol = inp->inp_socket->so_proto-> 1472 pr_protocol; 1473 pcb.ki_pflags = inp->inp_flags; 1474 1475 pcb.ki_sostate = inp->inp_socket->so_state; 1476 pcb.ki_prstate = inp->inp_state; 1477 if (proto == IPPROTO_TCP) { 1478 tp = intotcpcb(inp); 1479 pcb.ki_tstate = tp->t_state; 1480 pcb.ki_tflags = tp->t_flags; 1481 } 1482 1483 pcb.ki_pcbaddr = PTRTOUINT64(inp); 1484 pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb); 1485 pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket); 1486 1487 pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc; 1488 pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc; 1489 1490 in = satosin(&pcb.ki_src); 1491 in->sin_len = sizeof(*in); 1492 in->sin_family = pf; 1493 in->sin_port = inp->inp_lport; 1494 in->sin_addr = inp->inp_laddr; 1495 if (pcb.ki_prstate >= INP_CONNECTED) { 1496 in = satosin(&pcb.ki_dst); 1497 in->sin_len = sizeof(*in); 1498 in->sin_family = pf; 1499 in->sin_port = inp->inp_fport; 1500 in->sin_addr = inp->inp_faddr; 1501 } 1502 break; 1503 #endif 1504 #ifdef INET6 1505 case PF_INET6: 1506 pcb.ki_family = in6p->in6p_socket->so_proto-> 1507 pr_domain->dom_family; 1508 pcb.ki_type = in6p->in6p_socket->so_proto->pr_type; 1509 pcb.ki_protocol = in6p->in6p_socket->so_proto-> 1510 pr_protocol; 1511 pcb.ki_pflags = in6p->in6p_flags; 1512 1513 pcb.ki_sostate = in6p->in6p_socket->so_state; 1514 pcb.ki_prstate = in6p->in6p_state; 1515 if (proto == IPPROTO_TCP) { 1516 tp = in6totcpcb(in6p); 1517 pcb.ki_tstate = tp->t_state; 1518 pcb.ki_tflags = tp->t_flags; 1519 } 1520 1521 pcb.ki_pcbaddr = PTRTOUINT64(in6p); 1522 pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb); 1523 pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket); 1524 1525 pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc; 1526 pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc; 1527 1528 in6 = satosin6(&pcb.ki_src); 1529 in6->sin6_len = sizeof(*in6); 1530 in6->sin6_family = pf; 1531 in6->sin6_port = in6p->in6p_lport; 1532 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1533 in6->sin6_addr = in6p->in6p_laddr; 1534 in6->sin6_scope_id = 0; /* XXX? */ 1535 1536 if (pcb.ki_prstate >= IN6P_CONNECTED) { 1537 in6 = satosin6(&pcb.ki_dst); 1538 in6->sin6_len = sizeof(*in6); 1539 in6->sin6_family = pf; 1540 in6->sin6_port = in6p->in6p_fport; 1541 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1542 in6->sin6_addr = in6p->in6p_faddr; 1543 in6->sin6_scope_id = 0; /* XXX? */ 1544 } 1545 break; 1546 #endif 1547 } 1548 1549 if (len >= elem_size && elem_count > 0) { 1550 error = copyout(&pcb, dp, out_size); 1551 if (error) 1552 return (error); 1553 dp += elem_size; 1554 len -= elem_size; 1555 } 1556 if (elem_count > 0) { 1557 needed += elem_size; 1558 if (elem_count != INT_MAX) 1559 elem_count--; 1560 } 1561 } 1562 1563 *oldlenp = needed; 1564 if (oldp == NULL) 1565 *oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb); 1566 1567 return (error); 1568 } 1569 1570 static int 1571 sysctl_tcp_congctl(SYSCTLFN_ARGS) 1572 { 1573 struct sysctlnode node; 1574 int error, r; 1575 char newname[TCPCC_MAXLEN]; 1576 1577 strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1); 1578 1579 node = *rnode; 1580 node.sysctl_data = newname; 1581 node.sysctl_size = sizeof(newname); 1582 1583 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1584 1585 if (error || 1586 newp == NULL || 1587 strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0) 1588 return error; 1589 1590 if ((r = tcp_congctl_select(NULL, newname))) 1591 return r; 1592 1593 return error; 1594 } 1595 1596 static int 1597 sysctl_tcp_keep(SYSCTLFN_ARGS) 1598 { 1599 int error; 1600 u_int tmp; 1601 struct sysctlnode node; 1602 1603 node = *rnode; 1604 tmp = *(u_int *)rnode->sysctl_data; 1605 node.sysctl_data = &tmp; 1606 1607 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1608 if (error || newp == NULL) 1609 return error; 1610 1611 *(u_int *)rnode->sysctl_data = tmp; 1612 tcp_tcpcb_template(); /* update the template */ 1613 return 0; 1614 } 1615 1616 static void 1617 tcpstat_convert_to_user_cb(void *v1, void *v2, struct cpu_info *ci) 1618 { 1619 uint64_t *tcpsc = v1; 1620 uint64_t *tcps = v2; 1621 u_int i; 1622 1623 for (i = 0; i < TCP_NSTATS; i++) 1624 tcps[i] += tcpsc[i]; 1625 } 1626 1627 static void 1628 tcpstat_convert_to_user(uint64_t *tcps) 1629 { 1630 1631 memset(tcps, 0, sizeof(uint64_t) * TCP_NSTATS); 1632 percpu_foreach(tcpstat_percpu, tcpstat_convert_to_user_cb, tcps); 1633 } 1634 1635 static int 1636 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS) 1637 { 1638 struct sysctlnode node; 1639 uint64_t tcps[TCP_NSTATS]; 1640 1641 tcpstat_convert_to_user(tcps); 1642 node = *rnode; 1643 node.sysctl_data = tcps; 1644 node.sysctl_size = sizeof(tcps); 1645 return (sysctl_lookup(SYSCTLFN_CALL(&node))); 1646 } 1647 1648 /* 1649 * this (second stage) setup routine is a replacement for tcp_sysctl() 1650 * (which is currently used for ipv4 and ipv6) 1651 */ 1652 static void 1653 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname, 1654 const char *tcpname) 1655 { 1656 const struct sysctlnode *sack_node; 1657 const struct sysctlnode *abc_node; 1658 const struct sysctlnode *ecn_node; 1659 const struct sysctlnode *congctl_node; 1660 #ifdef TCP_DEBUG 1661 extern struct tcp_debug tcp_debug[TCP_NDEBUG]; 1662 extern int tcp_debx; 1663 #endif 1664 1665 sysctl_createv(clog, 0, NULL, NULL, 1666 CTLFLAG_PERMANENT, 1667 CTLTYPE_NODE, "net", NULL, 1668 NULL, 0, NULL, 0, 1669 CTL_NET, CTL_EOL); 1670 sysctl_createv(clog, 0, NULL, NULL, 1671 CTLFLAG_PERMANENT, 1672 CTLTYPE_NODE, pfname, NULL, 1673 NULL, 0, NULL, 0, 1674 CTL_NET, pf, CTL_EOL); 1675 sysctl_createv(clog, 0, NULL, NULL, 1676 CTLFLAG_PERMANENT, 1677 CTLTYPE_NODE, tcpname, 1678 SYSCTL_DESCR("TCP related settings"), 1679 NULL, 0, NULL, 0, 1680 CTL_NET, pf, IPPROTO_TCP, CTL_EOL); 1681 1682 sysctl_createv(clog, 0, NULL, NULL, 1683 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1684 CTLTYPE_INT, "rfc1323", 1685 SYSCTL_DESCR("Enable RFC1323 TCP extensions"), 1686 NULL, 0, &tcp_do_rfc1323, 0, 1687 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL); 1688 sysctl_createv(clog, 0, NULL, NULL, 1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1690 CTLTYPE_INT, "sendspace", 1691 SYSCTL_DESCR("Default TCP send buffer size"), 1692 NULL, 0, &tcp_sendspace, 0, 1693 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL); 1694 sysctl_createv(clog, 0, NULL, NULL, 1695 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1696 CTLTYPE_INT, "recvspace", 1697 SYSCTL_DESCR("Default TCP receive buffer size"), 1698 NULL, 0, &tcp_recvspace, 0, 1699 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL); 1700 sysctl_createv(clog, 0, NULL, NULL, 1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1702 CTLTYPE_INT, "mssdflt", 1703 SYSCTL_DESCR("Default maximum segment size"), 1704 sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0, 1705 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL); 1706 sysctl_createv(clog, 0, NULL, NULL, 1707 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1708 CTLTYPE_INT, "minmss", 1709 SYSCTL_DESCR("Lower limit for TCP maximum segment size"), 1710 NULL, 0, &tcp_minmss, 0, 1711 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1712 sysctl_createv(clog, 0, NULL, NULL, 1713 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1714 CTLTYPE_INT, "syn_cache_limit", 1715 SYSCTL_DESCR("Maximum number of entries in the TCP " 1716 "compressed state engine"), 1717 NULL, 0, &tcp_syn_cache_limit, 0, 1718 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT, 1719 CTL_EOL); 1720 sysctl_createv(clog, 0, NULL, NULL, 1721 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1722 CTLTYPE_INT, "syn_bucket_limit", 1723 SYSCTL_DESCR("Maximum number of entries per hash " 1724 "bucket in the TCP compressed state " 1725 "engine"), 1726 NULL, 0, &tcp_syn_bucket_limit, 0, 1727 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT, 1728 CTL_EOL); 1729 #if 0 /* obsoleted */ 1730 sysctl_createv(clog, 0, NULL, NULL, 1731 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1732 CTLTYPE_INT, "syn_cache_interval", 1733 SYSCTL_DESCR("TCP compressed state engine's timer interval"), 1734 NULL, 0, &tcp_syn_cache_interval, 0, 1735 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER, 1736 CTL_EOL); 1737 #endif 1738 sysctl_createv(clog, 0, NULL, NULL, 1739 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1740 CTLTYPE_INT, "init_win", 1741 SYSCTL_DESCR("Initial TCP congestion window"), 1742 NULL, 0, &tcp_init_win, 0, 1743 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL); 1744 sysctl_createv(clog, 0, NULL, NULL, 1745 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1746 CTLTYPE_INT, "mss_ifmtu", 1747 SYSCTL_DESCR("Use interface MTU for calculating MSS"), 1748 NULL, 0, &tcp_mss_ifmtu, 0, 1749 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL); 1750 sysctl_createv(clog, 0, NULL, &sack_node, 1751 CTLFLAG_PERMANENT, 1752 CTLTYPE_NODE, "sack", 1753 SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"), 1754 NULL, 0, NULL, 0, 1755 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL); 1756 1757 /* Congctl subtree */ 1758 sysctl_createv(clog, 0, NULL, &congctl_node, 1759 CTLFLAG_PERMANENT, 1760 CTLTYPE_NODE, "congctl", 1761 SYSCTL_DESCR("TCP Congestion Control"), 1762 NULL, 0, NULL, 0, 1763 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1764 sysctl_createv(clog, 0, &congctl_node, NULL, 1765 CTLFLAG_PERMANENT, 1766 CTLTYPE_STRING, "available", 1767 SYSCTL_DESCR("Available Congestion Control Mechanisms"), 1768 NULL, 0, &tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL); 1769 sysctl_createv(clog, 0, &congctl_node, NULL, 1770 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1771 CTLTYPE_STRING, "selected", 1772 SYSCTL_DESCR("Selected Congestion Control Mechanism"), 1773 sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN, 1774 CTL_CREATE, CTL_EOL); 1775 1776 sysctl_createv(clog, 0, NULL, NULL, 1777 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1778 CTLTYPE_INT, "win_scale", 1779 SYSCTL_DESCR("Use RFC1323 window scale options"), 1780 NULL, 0, &tcp_do_win_scale, 0, 1781 CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL); 1782 sysctl_createv(clog, 0, NULL, NULL, 1783 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1784 CTLTYPE_INT, "timestamps", 1785 SYSCTL_DESCR("Use RFC1323 time stamp options"), 1786 NULL, 0, &tcp_do_timestamps, 0, 1787 CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL); 1788 sysctl_createv(clog, 0, NULL, NULL, 1789 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1790 CTLTYPE_INT, "compat_42", 1791 SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"), 1792 NULL, 0, &tcp_compat_42, 0, 1793 CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL); 1794 sysctl_createv(clog, 0, NULL, NULL, 1795 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1796 CTLTYPE_INT, "cwm", 1797 SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window " 1798 "Monitoring"), 1799 NULL, 0, &tcp_cwm, 0, 1800 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL); 1801 sysctl_createv(clog, 0, NULL, NULL, 1802 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1803 CTLTYPE_INT, "cwm_burstsize", 1804 SYSCTL_DESCR("Congestion Window Monitoring allowed " 1805 "burst count in packets"), 1806 NULL, 0, &tcp_cwm_burstsize, 0, 1807 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE, 1808 CTL_EOL); 1809 sysctl_createv(clog, 0, NULL, NULL, 1810 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1811 CTLTYPE_INT, "ack_on_push", 1812 SYSCTL_DESCR("Immediately return ACK when PSH is " 1813 "received"), 1814 NULL, 0, &tcp_ack_on_push, 0, 1815 CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL); 1816 sysctl_createv(clog, 0, NULL, NULL, 1817 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1818 CTLTYPE_INT, "keepidle", 1819 SYSCTL_DESCR("Allowed connection idle ticks before a " 1820 "keepalive probe is sent"), 1821 sysctl_tcp_keep, 0, &tcp_keepidle, 0, 1822 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL); 1823 sysctl_createv(clog, 0, NULL, NULL, 1824 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1825 CTLTYPE_INT, "keepintvl", 1826 SYSCTL_DESCR("Ticks before next keepalive probe is sent"), 1827 sysctl_tcp_keep, 0, &tcp_keepintvl, 0, 1828 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL); 1829 sysctl_createv(clog, 0, NULL, NULL, 1830 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1831 CTLTYPE_INT, "keepcnt", 1832 SYSCTL_DESCR("Number of keepalive probes to send"), 1833 sysctl_tcp_keep, 0, &tcp_keepcnt, 0, 1834 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL); 1835 sysctl_createv(clog, 0, NULL, NULL, 1836 CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE, 1837 CTLTYPE_INT, "slowhz", 1838 SYSCTL_DESCR("Keepalive ticks per second"), 1839 NULL, PR_SLOWHZ, NULL, 0, 1840 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL); 1841 sysctl_createv(clog, 0, NULL, NULL, 1842 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1843 CTLTYPE_INT, "log_refused", 1844 SYSCTL_DESCR("Log refused TCP connections"), 1845 NULL, 0, &tcp_log_refused, 0, 1846 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL); 1847 #if 0 /* obsoleted */ 1848 sysctl_createv(clog, 0, NULL, NULL, 1849 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1850 CTLTYPE_INT, "rstratelimit", NULL, 1851 NULL, 0, &tcp_rst_ratelim, 0, 1852 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL); 1853 #endif 1854 sysctl_createv(clog, 0, NULL, NULL, 1855 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1856 CTLTYPE_INT, "rstppslimit", 1857 SYSCTL_DESCR("Maximum number of RST packets to send " 1858 "per second"), 1859 NULL, 0, &tcp_rst_ppslim, 0, 1860 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL); 1861 sysctl_createv(clog, 0, NULL, NULL, 1862 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1863 CTLTYPE_INT, "delack_ticks", 1864 SYSCTL_DESCR("Number of ticks to delay sending an ACK"), 1865 NULL, 0, &tcp_delack_ticks, 0, 1866 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL); 1867 sysctl_createv(clog, 0, NULL, NULL, 1868 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1869 CTLTYPE_INT, "init_win_local", 1870 SYSCTL_DESCR("Initial TCP window size (in segments)"), 1871 NULL, 0, &tcp_init_win_local, 0, 1872 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL, 1873 CTL_EOL); 1874 sysctl_createv(clog, 0, NULL, NULL, 1875 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1876 CTLTYPE_STRUCT, "ident", 1877 SYSCTL_DESCR("RFC1413 Identification Protocol lookups"), 1878 sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t), 1879 CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL); 1880 sysctl_createv(clog, 0, NULL, NULL, 1881 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1882 CTLTYPE_INT, "do_loopback_cksum", 1883 SYSCTL_DESCR("Perform TCP checksum on loopback"), 1884 NULL, 0, &tcp_do_loopback_cksum, 0, 1885 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM, 1886 CTL_EOL); 1887 sysctl_createv(clog, 0, NULL, NULL, 1888 CTLFLAG_PERMANENT, 1889 CTLTYPE_STRUCT, "pcblist", 1890 SYSCTL_DESCR("TCP protocol control block list"), 1891 sysctl_inpcblist, 0, &tcbtable, 0, 1892 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 1893 CTL_EOL); 1894 sysctl_createv(clog, 0, NULL, NULL, 1895 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1896 CTLTYPE_INT, "keepinit", 1897 SYSCTL_DESCR("Ticks before initial tcp connection times out"), 1898 sysctl_tcp_keep, 0, &tcp_keepinit, 0, 1899 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1900 1901 /* TCP socket buffers auto-sizing nodes */ 1902 sysctl_createv(clog, 0, NULL, NULL, 1903 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1904 CTLTYPE_INT, "recvbuf_auto", 1905 SYSCTL_DESCR("Enable automatic receive " 1906 "buffer sizing (experimental)"), 1907 NULL, 0, &tcp_do_autorcvbuf, 0, 1908 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1909 sysctl_createv(clog, 0, NULL, NULL, 1910 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1911 CTLTYPE_INT, "recvbuf_inc", 1912 SYSCTL_DESCR("Incrementor step size of " 1913 "automatic receive buffer"), 1914 NULL, 0, &tcp_autorcvbuf_inc, 0, 1915 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1916 sysctl_createv(clog, 0, NULL, NULL, 1917 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1918 CTLTYPE_INT, "recvbuf_max", 1919 SYSCTL_DESCR("Max size of automatic receive buffer"), 1920 NULL, 0, &tcp_autorcvbuf_max, 0, 1921 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1922 1923 sysctl_createv(clog, 0, NULL, NULL, 1924 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1925 CTLTYPE_INT, "sendbuf_auto", 1926 SYSCTL_DESCR("Enable automatic send " 1927 "buffer sizing (experimental)"), 1928 NULL, 0, &tcp_do_autosndbuf, 0, 1929 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1930 sysctl_createv(clog, 0, NULL, NULL, 1931 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1932 CTLTYPE_INT, "sendbuf_inc", 1933 SYSCTL_DESCR("Incrementor step size of " 1934 "automatic send buffer"), 1935 NULL, 0, &tcp_autosndbuf_inc, 0, 1936 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1937 sysctl_createv(clog, 0, NULL, NULL, 1938 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1939 CTLTYPE_INT, "sendbuf_max", 1940 SYSCTL_DESCR("Max size of automatic send buffer"), 1941 NULL, 0, &tcp_autosndbuf_max, 0, 1942 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1943 1944 /* ECN subtree */ 1945 sysctl_createv(clog, 0, NULL, &ecn_node, 1946 CTLFLAG_PERMANENT, 1947 CTLTYPE_NODE, "ecn", 1948 SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"), 1949 NULL, 0, NULL, 0, 1950 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1951 sysctl_createv(clog, 0, &ecn_node, NULL, 1952 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1953 CTLTYPE_INT, "enable", 1954 SYSCTL_DESCR("Enable TCP Explicit Congestion " 1955 "Notification"), 1956 NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL); 1957 sysctl_createv(clog, 0, &ecn_node, NULL, 1958 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1959 CTLTYPE_INT, "maxretries", 1960 SYSCTL_DESCR("Number of times to retry ECN setup " 1961 "before disabling ECN on the connection"), 1962 NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL); 1963 1964 /* SACK gets it's own little subtree. */ 1965 sysctl_createv(clog, 0, NULL, &sack_node, 1966 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1967 CTLTYPE_INT, "enable", 1968 SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"), 1969 NULL, 0, &tcp_do_sack, 0, 1970 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1971 sysctl_createv(clog, 0, NULL, &sack_node, 1972 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1973 CTLTYPE_INT, "maxholes", 1974 SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"), 1975 NULL, 0, &tcp_sack_tp_maxholes, 0, 1976 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1977 sysctl_createv(clog, 0, NULL, &sack_node, 1978 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1979 CTLTYPE_INT, "globalmaxholes", 1980 SYSCTL_DESCR("Global maximum number of TCP SACK holes"), 1981 NULL, 0, &tcp_sack_globalmaxholes, 0, 1982 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1983 sysctl_createv(clog, 0, NULL, &sack_node, 1984 CTLFLAG_PERMANENT, 1985 CTLTYPE_INT, "globalholes", 1986 SYSCTL_DESCR("Global number of TCP SACK holes"), 1987 NULL, 0, &tcp_sack_globalholes, 0, 1988 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1989 1990 sysctl_createv(clog, 0, NULL, NULL, 1991 CTLFLAG_PERMANENT, 1992 CTLTYPE_STRUCT, "stats", 1993 SYSCTL_DESCR("TCP statistics"), 1994 sysctl_net_inet_tcp_stats, 0, NULL, 0, 1995 CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS, 1996 CTL_EOL); 1997 #ifdef TCP_DEBUG 1998 sysctl_createv(clog, 0, NULL, NULL, 1999 CTLFLAG_PERMANENT, 2000 CTLTYPE_STRUCT, "debug", 2001 SYSCTL_DESCR("TCP sockets debug information"), 2002 NULL, 0, &tcp_debug, sizeof(tcp_debug), 2003 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG, 2004 CTL_EOL); 2005 sysctl_createv(clog, 0, NULL, NULL, 2006 CTLFLAG_PERMANENT, 2007 CTLTYPE_INT, "debx", 2008 SYSCTL_DESCR("Number of TCP debug sockets messages"), 2009 NULL, 0, &tcp_debx, sizeof(tcp_debx), 2010 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX, 2011 CTL_EOL); 2012 #endif 2013 sysctl_createv(clog, 0, NULL, NULL, 2014 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2015 CTLTYPE_STRUCT, "drop", 2016 SYSCTL_DESCR("TCP drop connection"), 2017 sysctl_net_inet_tcp_drop, 0, NULL, 0, 2018 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL); 2019 #if NRND > 0 2020 sysctl_createv(clog, 0, NULL, NULL, 2021 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2022 CTLTYPE_INT, "iss_hash", 2023 SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic " 2024 "hash computation"), 2025 NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948), 2026 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 2027 CTL_EOL); 2028 #endif 2029 2030 /* ABC subtree */ 2031 2032 sysctl_createv(clog, 0, NULL, &abc_node, 2033 CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc", 2034 SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"), 2035 NULL, 0, NULL, 0, 2036 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2037 sysctl_createv(clog, 0, &abc_node, NULL, 2038 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2039 CTLTYPE_INT, "enable", 2040 SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"), 2041 NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL); 2042 sysctl_createv(clog, 0, &abc_node, NULL, 2043 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2044 CTLTYPE_INT, "aggressive", 2045 SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"), 2046 NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL); 2047 } 2048 2049 /* 2050 * Sysctl for tcp variables. 2051 */ 2052 #ifdef INET 2053 SYSCTL_SETUP(sysctl_net_inet_tcp_setup, "sysctl net.inet.tcp subtree setup") 2054 { 2055 2056 sysctl_net_inet_tcp_setup2(clog, PF_INET, "inet", "tcp"); 2057 } 2058 #endif /* INET */ 2059 2060 #ifdef INET6 2061 SYSCTL_SETUP(sysctl_net_inet6_tcp6_setup, "sysctl net.inet6.tcp6 subtree setup") 2062 { 2063 2064 sysctl_net_inet_tcp_setup2(clog, PF_INET6, "inet6", "tcp6"); 2065 } 2066 #endif /* INET6 */ 2067